This is a guest post by John Church. John is an oil and gas professional and has spent some 30 years working for one of the biggest oil companies in the world.
The key uncertainties in this epidemic are how far and fast the infection has spread, and what the mortality is. There are lots of other variables and uncertainties about who gets affected (sex, ethnicity, BMI, BCG or not, whether being infected grants immunity thereafter, etc.). But let’s just use the information we have to try and work out approximate averages for infection spread and mortality (age dependent).
Based on the Diamond Princess cruise ship (DP) and early Italian hospital info we can deduce a few things:
1) No one under 40 seems to get seriously ill, if any symptoms at all. Yes, there have been a few highlighted cases in the media, but the numbers are very, very small.
2) We know from the DP data that there were 711 confirmed infections and ~20% required hospital. We know that this resulted in 12 or 13 deaths.
3) We know that mainly its people in their 60s and 70s that go on cruises. The age distribution on the DP infections is below.
4) We know that chance of hospitalisation and also mortality steadily increases with age.
5) We know from hospitalisations in Italy that ~14% of people above 80, and ~ 7% of people in their 70’s were dying. Probably higher than it might have been if they hadn’t run out of critical capacity, but let’s stick with these numbers.
6) About 7% of hospitalised patient deaths in the UK are under 60.
7) We know the population distribution in the UK:
>80 3.3mln ~5%
70-80 5.5mln ~8%
60-70 7.0mln ~11%
50-60 9.0mln ~13%
40-50 9.1mln ~14%
<40 32.8mln ~49%
Total 66.7mln 100%
The Diamond Princess (DP)
Using the age distribution of the passengers and looking at the observations (hospitalisations and fatalities), it is possible to create an age dependent infection model. On the Diamond Princess there were 711 cases and 12 deaths. Approximately 20% of cases required hospitalisation. The approx. age distribution of cases was >80 = 9%, 70-80 = 38%, 60-70 = 29%, 50-60 = 9%, 40-50 = 4% and <40 = 11%. A model which gives a good match to this data, fits the Italian hospital mortality data and honours the basic observations is below :
Age % hospitalised Hospital death rate Mortality (all infections)
>80 ~ 40% ~14 % 5.6%
70-80 ~ 30% ~ 7 % 2.1%
60-70 ~ 20% ~ 4 % 0.8%
50-60 ~ 10% ~ 2 % 0.2%
40-50 ~ 1% ~1 % 0%
This needs to be seen as a working model, but based on some solid observations. It can be adapted as more information come in.
So if this scenario was correct, what would it mean in the UK if everybody caught the disease? Or if we limit the spread until 60% herd immunity ?
Age Calculation Deaths % Deaths (if assume 60% herd)
>80 3.3 x 5.6% 185k 49% 111K
70-80 5.5 x 2.1% 116k 31% 69k
60-70 7.0 x 0.8% 56k 15% 34k
50-60 9.0 x 0.2% 18k 5% 11k
40-50 9.1 x 0.01% 1k 0% <1k
Total 375k 225k
So this would mean over several waves of infection, without vaccine and under current treatments we could have ~375k deaths if everyone caught it, but might expect to have a maximum of ~225k deaths assuming up to 60% of the population gets infected.
So what does this mean for average mortality and infection spread in the UK?
Well, if everyone caught it and we had 375k deaths, that means an average mortality of ~0.5%. About 5x that of common flu. Does that seem reasonable ? I see this as a bit of a worst case as it’s based primarily on the Italian data which may have had higher mortality due to the fact they exceeded their critical care capacity, and it also assumes we don’t learn which treatments are most effective as we progress.
What about spread of infection in the UK?
It looks as though this current wave of deaths will be ~ 35k, if we assume we have just passed the peak and we assume approximately one-third deaths before the peak and two-thirds after (as per Italy and Spain apparent distributions). In that case we would have already infected 35000/0.5% = about 7 million people (approximately 10% of population). But we also need to factor in that there are lots of deaths happening outside hospital (beyond the 35k). Analysing Holland vs Belgium (where they count all deaths, not just hospitals), it appears as though there are approximately double the number of deaths in Belgium (current rates are 202 deaths/million in Holland and 445 deaths/million in Belgium, according to the Times 18/4/20). And recent reports suggest 7500 care-home deaths, which would result in a 50% increase compared to hospitalised numbers. If we assume an additional 50%, this would result in approximately 50 000 deaths in the UK as part of this first wave, and that means 50000/0.5% = 10 million infected already (~15% of population).
So how about the geographic distribution?
We should assume a concentration in urban areas. I read somewhere today that a third of all UK deaths come from London. If that is the case then we have >12 000 deaths (1/3 of 35,000) in London hospitals in this wave, with maybe 18 000 total if we assume care home deaths result in a 50% increase to get the real number. With a mortality of 0.5%, this would mean some 3.5 million people in London have been infected, approx. 40% of the population of 9 million. So more infections in big urban areas but offset by lower rates elsewhere. As might be expected. It would also explain why so many famous people who live and work in London have been infected. Clearly there has been very widespread infection in the city.
What about spreadrate?
The DP timeline showed a two-week period between when a Covid-positive passenger embarked the liner in Yokohama on Jan 20th and Feb 4th when the ship was placed in quarantine. How often do we need to double the cases to get from 1 to 711 ? Well, it turn out that this needs a doubling rate every 1.5 days. Or the same passenger managed to infect 700 other people. Or a combination of superspreading in a confined environment and also a rapid spread rate. Either way, in a confined environment the spreadrate was very fast and efficient. Regarding the UK situation, given that the first cases came into the UK in early/mid January from China (returning students) and also from European ski resorts, and we didn’t enter lockdown until March 23rd, there is more than enough time for a very large number of people to get infected. There are 10 weeks between mid -January and March 23rd (lockdown day). If we assume infections double every 3 days, that’s 23 ‘doublings’ and results in over 8 million cases. This is in line with the DP and Italian data which predicts between 7 and 10 million cases as above. A study done by Oxford University stated a high case scenario of 50% infections (> 33 million people), but this is probably unlikely as this would lead to much lower mortality than shown by the DP and Italian data.
So what does this mean for the future?
Once the current (first) wave has diminished, and restrictions are eased, there will be a continuation of infections. However, if we already have 10-15% infection rates (with up to 40% in urban areas) the second wave will be slower and lower. In urban areas the higher previous infections will mean it spreads slowly, and in less populated areas the lower population density will mean slower spread. So we will see nothing like the firestorm we have seen in the first wave. If we look at the 10-week timing from arrival in UK (mid-January) to lockdown for the first wave, we could predict another (second) wave in about 2-3 months from the ending of restrictions. So assuming we relax restrictions in early May, then we could predict a second wave during August and September or sometime beyond. Based on 50 000 deaths due to the first wave (hospital and non-hospital), by the end of the year we might expect an additional 25 000 to 50 000 deaths. But from a purely forensic point of view, this will increasingly manageable as each wave will be less of a problem than the previous.
What should we be doing?
I’m not an epidemiologist, but it would seem to me, based on the calculations above, that this is very infectious and spreads very fast (7-10 million cases in the UK already), has an average mortality of ~ 0.5%, doesn’t damage younger people at all, is relatively low risk for anyone less than 60, and the NHS has been shown to cope with the first and largest wave, so we should protect older at-risk people through isolation as best we can and let this thing run its course as fast as possible. Once 60% have had it, the risk diminishes for everyone. There is no point applying a containment strategy, as you suffer the economic consequences but gain none of the benefits of increased population immunity, because everyone will get infected anyway. This can be seen from the Singapore data right now. Australia and NZ will be test cases when they re-open their economies. They have not even had a ‘first wave’, so will there be a surge of cases 2-3 months after they ease restrictions? And we should continue to monitor Sweden who have not put in place stringent isolation for the majority of their population: if there is no escalation to an out-of-control situation, it tells you we should just look after older people but let it rip through everyone else. ASAP.
So, there are 3 key reasons to get this through the population as soon as possible,
1) Economic: the faster this is over the less the economic downturn. We should have all younger people working now, but getting the full workforce back as soon as is practical.
2) Societal: minimise the time older generations and those ‘at risk’ must remain in isolation. It is not healthy, physically, mentally and from a societal perspective, to have this in place and the negative impacts are not fully known.
3) Medical: being exposed to this virus sooner means that everyone is younger when it happens. Given that this is a virus which gets worse with age, it makes sense to be exposed to it as early as possible.
What if it turns out there is no immunity and people can get re-infected?
Well, we are just going to have to live with it until a vaccine is found. Mortality will decline even in older people as we understand better how to treat people. The problem will just fade away and become part of life and death.
How does all this compare to ‘business as usual’?
An expected death toll of 225k overall (50k in this first wave) sounds like a lot of people, but we must remember that this is only 5 months of normal deaths in the UK which occur at around 50k per month. According to hospital reports, 90% of those dying have serious co-morbidities, particularly cardio-vascular disease and high blood pressure, and many would have died shortly anyway.
According to Professor Johan Giesecke (pers. comm.), who is a Swedish epidemiologist advising the Swedish government and also the WHO, in order to ascertain the ‘lethality’ of the epidemic we need to remove these cases. Doing so (removing 90% of the cases) would reduce the 0.5% mortality figure to a ‘lethality’ of around 0.1%, which is what a normal healthy population should consider. This is approximately the same as a bad flu epidemic. We will only know this when we review the whole of 2020 and see how the total UK deaths compare (month by month) to long term averages. It is these excess deaths which ultimately determine the true nature of the epidemic.
Whatever else we do, we should continue to look at the Swedish situation. If they have managed their pandemic successfully without population-wide enforced isolations we should definitely not put all our healthy young people into lockdown again.